Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007755
A. Banerjee, Joydeep Ghosh
This paper derives three competitive learning mechanisms from first principles to obtain clusters of comparable sizes when both inputs and representatives are normalized. These mechanisms are very effective in achieving balanced grouping of inputs in high dimensional spaces as illustrated by experimental results on clustering two popular text data sets in 26,099 and 21,839 dimensional spaces, respectively.
{"title":"Frequency sensitive competitive learning for clustering on high-dimensional hyperspheres","authors":"A. Banerjee, Joydeep Ghosh","doi":"10.1109/IJCNN.2002.1007755","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007755","url":null,"abstract":"This paper derives three competitive learning mechanisms from first principles to obtain clusters of comparable sizes when both inputs and representatives are normalized. These mechanisms are very effective in achieving balanced grouping of inputs in high dimensional spaces as illustrated by experimental results on clustering two popular text data sets in 26,099 and 21,839 dimensional spaces, respectively.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115688328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007509
R. Neubert, O. Gorlitz, W. Benn, T. Teich
Presents the idea and first results of using GNG networks for hierarchical cluster analysis in order to create index structures for data management systems. It describes the creation procedure of a multi-dimensional index structure, the Intelligent Cluster Index (ICIx). In particular critical design decisions and tradeoffs between index efficiency and the neural network's clustering solution are discussed.
{"title":"Obstacles for neural network application in the ICIx database index","authors":"R. Neubert, O. Gorlitz, W. Benn, T. Teich","doi":"10.1109/IJCNN.2002.1007509","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007509","url":null,"abstract":"Presents the idea and first results of using GNG networks for hierarchical cluster analysis in order to create index structures for data management systems. It describes the creation procedure of a multi-dimensional index structure, the Intelligent Cluster Index (ICIx). In particular critical design decisions and tradeoffs between index efficiency and the neural network's clustering solution are discussed.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"114 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123115286","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007516
S. Vishwanathan, M. Narasimha Murty
We present a fast iterative algorithm for identifying the support vectors of a given set of points. Our algorithm works by maintaining a candidate support vector set. It uses a greedy approach to pick points for inclusion in the candidate set. When the addition of a point to the candidate set is blocked because of other points already present in the set, we use a backtracking approach to prune away such points. To speed up convergence we initialize our algorithm with the nearest pair of points from opposite classes. We then use an optimization based approach to increase or prune the candidate support vector set. The algorithm makes repeated passes over the data to satisfy the KKT constraints. The memory requirements of our algorithm scale as O(|SI|/sup 2/) in the average case, where |S| is the size of the support vector set. We show that the algorithm is extremely competitive as compared to other conventional iterative algorithms like SMO and the NPA. We present results on a variety of real life datasets to validate our claims.
{"title":"SSVM: a simple SVM algorithm","authors":"S. Vishwanathan, M. Narasimha Murty","doi":"10.1109/IJCNN.2002.1007516","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007516","url":null,"abstract":"We present a fast iterative algorithm for identifying the support vectors of a given set of points. Our algorithm works by maintaining a candidate support vector set. It uses a greedy approach to pick points for inclusion in the candidate set. When the addition of a point to the candidate set is blocked because of other points already present in the set, we use a backtracking approach to prune away such points. To speed up convergence we initialize our algorithm with the nearest pair of points from opposite classes. We then use an optimization based approach to increase or prune the candidate support vector set. The algorithm makes repeated passes over the data to satisfy the KKT constraints. The memory requirements of our algorithm scale as O(|SI|/sup 2/) in the average case, where |S| is the size of the support vector set. We show that the algorithm is extremely competitive as compared to other conventional iterative algorithms like SMO and the NPA. We present results on a variety of real life datasets to validate our claims.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"176 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123050742","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007794
G. Yen, P. Lima
In this paper a hierarchical architecture that combines a high degree of reconfigurability and long-term memory is proposed as a fault tolerant control algorithm for complex nonlinear systems. Dual heuristic programming (DHP) is used for adapting to faults as they occur for the first time in an effort to prevent the build up of a general failure and also as tuning device after switching to a known scenario. A dynamical database, initialized with as much information of the plant as available, oversees the DHP controller. The decisions of which environments to record, when to intervene and where to switch are autonomously taken based on specifically designed quality indexes. The results of the application of the complete algorithm to a proof-of-the-concept numerical example help to illustrate the fine interrelations between each of its subsystems.
{"title":"Adaptive critic fault tolerant control using dual heuristic programming","authors":"G. Yen, P. Lima","doi":"10.1109/IJCNN.2002.1007794","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007794","url":null,"abstract":"In this paper a hierarchical architecture that combines a high degree of reconfigurability and long-term memory is proposed as a fault tolerant control algorithm for complex nonlinear systems. Dual heuristic programming (DHP) is used for adapting to faults as they occur for the first time in an effort to prevent the build up of a general failure and also as tuning device after switching to a known scenario. A dynamical database, initialized with as much information of the plant as available, oversees the DHP controller. The decisions of which environments to record, when to intervene and where to switch are autonomously taken based on specifically designed quality indexes. The results of the application of the complete algorithm to a proof-of-the-concept numerical example help to illustrate the fine interrelations between each of its subsystems.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"125 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123070689","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007803
N.F. Shakirova, L.N. Stolyarov, E. Stolyarova
We suggest a model of the asynchronous behavior of speculators on the capital market in the form of a Petri net. To analyze Petri nets from the viewpoint of independent transitions we suggest an original method, which makes it possible to construct a logical scheme of the operation of a Petri net. Auxiliary constructions of the method are also used for the definition of notions of a sequential-parallel scenario, a pattern of such a scenario, and the predictability or a scenario along a chain or patterns.
{"title":"Petri nets for modeling the behavior of speculators","authors":"N.F. Shakirova, L.N. Stolyarov, E. Stolyarova","doi":"10.1109/IJCNN.2002.1007803","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007803","url":null,"abstract":"We suggest a model of the asynchronous behavior of speculators on the capital market in the form of a Petri net. To analyze Petri nets from the viewpoint of independent transitions we suggest an original method, which makes it possible to construct a logical scheme of the operation of a Petri net. Auxiliary constructions of the method are also used for the definition of notions of a sequential-parallel scenario, a pattern of such a scenario, and the predictability or a scenario along a chain or patterns.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116657764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007652
O. Duran, K. Althoefer, L. Seneviratne
The focus of the research presented here is on the automated assessment of sewer pipe conditions using a laser-based sensor. The proposed method involves image and data processing algorithms categorising signals acquired from the internal pipe surface. Fault identification is carried out using a neural network. Experimental results are presented.
{"title":"Automated sewer inspection using image processing and a neural classifier","authors":"O. Duran, K. Althoefer, L. Seneviratne","doi":"10.1109/IJCNN.2002.1007652","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007652","url":null,"abstract":"The focus of the research presented here is on the automated assessment of sewer pipe conditions using a laser-based sensor. The proposed method involves image and data processing algorithms categorising signals acquired from the internal pipe surface. Fault identification is carried out using a neural network. Experimental results are presented.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"138 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121114946","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007722
G. Backfried, R. Rainoldi, J. Riedler
We present experiments on automatic language identification in the broadcast news domain. Because of the inherent diversity of news broadcasts, speech is extracted from the raw audio data by means of phone-level decoding using broad classes of phonemes. Training and testing was performed on recordings of German, English, Spanish and French news shows from a variety of European TV channels. Each language is characterized by a Gaussian mixture model solely created from corresponding acoustic features. The overall average error rate on speech segments is 16.32%. The current system disregards (almost) any kind of linguistic information; however, it is therefore easily extensible to new languages.
{"title":"Automatic language identification in broadcast news","authors":"G. Backfried, R. Rainoldi, J. Riedler","doi":"10.1109/IJCNN.2002.1007722","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007722","url":null,"abstract":"We present experiments on automatic language identification in the broadcast news domain. Because of the inherent diversity of news broadcasts, speech is extracted from the raw audio data by means of phone-level decoding using broad classes of phonemes. Training and testing was performed on recordings of German, English, Spanish and French news shows from a variety of European TV channels. Each language is characterized by a Gaussian mixture model solely created from corresponding acoustic features. The overall average error rate on speech segments is 16.32%. The current system disregards (almost) any kind of linguistic information; however, it is therefore easily extensible to new languages.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"54 3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127215081","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1005457
Shahidul Pramanik, Robert Kozma, Dipankar Dasgupta
The germinal center (GC) is a functional module positioned in strategic locations of the lymphatic network in the animal body, which is known to play an important role in the immune response. Its formation and function can be explained and analyzed from a computational point of view using neural network technology. The objective of the paper is to model GC organization in terms of NN architecture and dynamics. A cascade of three Hopfield networks along with the Hebbian learning principle is used in a data classification problem where the connection matrices determine the local and global feedback as well as the propagation from one state to another in the network.
生发中心(germinal center, GC)是一种功能模块，位于动物体内淋巴网络的重要位置，在免疫反应中发挥重要作用。它的形成和功能可以用神经网络技术从计算的角度来解释和分析。本文的目的是根据神经网络的结构和动态对GC组织进行建模。三个Hopfield网络的级联以及Hebbian学习原理用于数据分类问题，其中连接矩阵决定了局部和全局反馈以及网络中从一种状态到另一种状态的传播。
{"title":"Dynamical neuro-representation of an immune model and its application for data classification","authors":"Shahidul Pramanik, Robert Kozma, Dipankar Dasgupta","doi":"10.1109/IJCNN.2002.1005457","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1005457","url":null,"abstract":"The germinal center (GC) is a functional module positioned in strategic locations of the lymphatic network in the animal body, which is known to play an important role in the immune response. Its formation and function can be explained and analyzed from a computational point of view using neural network technology. The objective of the paper is to model GC organization in terms of NN architecture and dynamics. A cascade of three Hopfield networks along with the Hebbian learning principle is used in a data classification problem where the connection matrices determine the local and global feedback as well as the propagation from one state to another in the network.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127298901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007464
G. Plett, H. Bottrich
Adaptive inverse control (AIC) uses three adaptive filters: plant model, controller and disturbance canceler. Methods are known for quick and efficient training of these filters if the plant is linear; however, known methods for nonlinear AIC learn very slowly. This paper modifies the standard nonlinear AIC learning methods (based on real-time recurrent learning) using the dynamic-decoupled-extended Kalman-filter (DDEKF). The training becomes significantly faster.
{"title":"DDEKF learning for fast nonlinear adaptive inverse control","authors":"G. Plett, H. Bottrich","doi":"10.1109/IJCNN.2002.1007464","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007464","url":null,"abstract":"Adaptive inverse control (AIC) uses three adaptive filters: plant model, controller and disturbance canceler. Methods are known for quick and efficient training of these filters if the plant is linear; however, known methods for nonlinear AIC learn very slowly. This paper modifies the standard nonlinear AIC learning methods (based on real-time recurrent learning) using the dynamic-decoupled-extended Kalman-filter (DDEKF). The training becomes significantly faster.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127381510","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2002-08-07DOI: 10.1109/IJCNN.2002.1007454
E. Saatci, V. Tavsanoglu
This paper presents a multi-scale character recognition system consisting of three single-scale recognition systems. The system uses a filter bank of Gabor-type filters implemented by a cellular neural network (CNN). Based on a test set of 26 test characters acting as template and a set consisting of four subsets of 26 unknown handwritten test characters, a maximum 96% and an average 93% correct recognition is provided. This is a considerable improvement over the performance of existing single-scale recognition systems.
{"title":"Multiscale handwritten character recognition using CNN image filters","authors":"E. Saatci, V. Tavsanoglu","doi":"10.1109/IJCNN.2002.1007454","DOIUrl":"https://doi.org/10.1109/IJCNN.2002.1007454","url":null,"abstract":"This paper presents a multi-scale character recognition system consisting of three single-scale recognition systems. The system uses a filter bank of Gabor-type filters implemented by a cellular neural network (CNN). Based on a test set of 26 test characters acting as template and a set consisting of four subsets of 26 unknown handwritten test characters, a maximum 96% and an average 93% correct recognition is provided. This is a considerable improvement over the performance of existing single-scale recognition systems.","PeriodicalId":382771,"journal":{"name":"Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290)","volume":"192 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2002-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125846898","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：